Professional Documents
Culture Documents
Energy Policy
journal homepage: www.elsevier.com/locate/enpol
H I G H L I G H T S
c We model a wind and solar based European power system with storage and balancing.
c We find that storage needs peaks when average renewable generation matches load.
c We find strong synergetic effects when combining storage and balancing.
c We study the effects of a storage capable of storing 6 h average use.
c We find a realisable fully renewable scenario based on wind, solar and hydro power.
a r t i c l e i n f o a b s t r a c t
Article history: Through a parametric time-series analysis of 8 years of hourly data, we quantify the storage size and
Received 24 March 2012 balancing energy needs for highly and fully renewable European power systems for different levels and
Accepted 4 September 2012 mixes of wind and solar energy. By applying a dispatch strategy that minimizes the balancing energy
Available online 25 September 2012
needs for a given storage size, the interplay between storage and balancing is quantified, providing a
Keywords: hard upper limit on their synergy. An efficient but relatively small storage reduces balancing energy
Energy system design needs significantly due to its influence on intra-day mismatches. Furthermore, we show that combined
Large-scale integration with a low-efficiency hydrogen storage and a level of balancing equal to what is today provided by
Storage storage lakes, it is sufficient to meet the European electricity demand in a fully renewable power
system where the average power generation from combined wind and solar exceeds the demand by
only a few percent.
& 2012 Elsevier Ltd. All rights reserved.
1. Introduction combination with either only storage (Heide et al., 2010) or only
balancing (Heide et al., 2011) requires a very large amount of excess
A fully renewable pan-European power system will depend on a generation in order to be technically feasible. Here, we study the
large share of non-dispatchable, weather dependent sources, intermediate and more realistic scenarios where power generation
primarily wind and solar power (Czisch, 2005; Jacobson and from the two weather-driven variable renewable energy (VRE)
Delucchi, 2011). The optimal ratio between and necessary amount sources is backed up by specific combinations of storage and
of wind and solar power depends on storage and balancing resources balancing. We identify a class of realistic and feasible scenarios for
(Heide et al., 2010, 2011; Hedegaard and Meibom, 2012), transmis- building a fully or partially renewable pan-European power system.
sion (Czisch et al., 2007; Kempton et al., 2010; Schaber et al., 2012a,b), The main point of our paper is to outline what is possible in a
and the characteristics of the climate (Widén, 2011; Aboumahboub wind and solar based European power system with storage and
et al., 2010) and load (Yao and Steemers, 2005). Based on balancing systems. The power capacities of the storage and
meteorological data, it was shown that even with unlimited trans- balancing facilities are not determined; this would require a more
mission within Europe, a scenario with only wind and solar power in complex modeling with explicit inclusion of power transmission
(Rodriguez et al., 2012). We focus on wind and solar power and
assume no bottlenecks in the power grid, employ an optimal
n
Corresponding author at: Center for Mathematical Sciences, Technische
storage dispatch strategy and ignore storage charge and discharge
Universität München, 85748 Garching, Germany. capacities and economic aspects. The incentives for doing so are
E-mail addresses: mgr@imf.au.dk, mortengrud@gmail.com (M.G. Rasmussen). closely related and at least threefold.
0301-4215/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.enpol.2012.09.009
M.G. Rasmussen et al. / Energy Policy 51 (2012) 642–651 643
VRE2 generation factor. An excerpt of the mismatch time series DðtÞ operation works. Three other versions satisfying the minimum EB
for an average VRE generation factor g ¼ 1:00 and a mix of condition are presented in the appendix.
aW ¼ 0:60 together with the corresponding excerpt of the load The version presented here can be summed up as a ‘‘storage
and wind and solar power time series can be seen in Fig. 1. first’’ strategy, in the sense that any deficits are first covered with
storage unless it runs empty, and any excess generation is stored
2.3. Balancing in the storage, unless the storage gets full. No limits are imposed
on the charge and discharge capacities. Conversion losses in and
As our focus is on wind and solar based, fully renewable power out of the storage are modeled by storage efficiencies Zin and Zout . For
systems, we consider any dispatchable, additional power source example, for hydrogen storage, the efficiencies are approximately
to be balancing (excluding storage). Thus, when no storage is 0.60 in both directions (Beaudin et al., 2010; Kruse et al., 2002),
present, the balancing time series is simply the residual load, or leading to a round-trip efficiency of 0.36.
minus the negative values of DðtÞ A storage is thus characterized by three parameters: Zin , Zout
and CS. The storage filling level time series H(t) with a constrained
BðtÞ ¼ D ðtÞ ðno storageÞ, ð3Þ
storage size CS is given by
corresponding to the red areas in the lower panel of Fig. 1. In this 8 ~ ðtÞ 4C S ,
paper, we will mainly focus on one property of the balancing time < CS
> for Hðt1Þ þ D
series, namely its average value (which we divide by the average HðtÞ ¼ 0 ~ ðtÞ o0,
for Hðt1Þ þ D ð5Þ
>
:
load to get a dimensionless parameter) ~ ðtÞ
Hðt1Þ þ D otherwise,
EB ¼ /BS=/LS,
~ is given by the equation
where D
EB measures the share of the energy that must be provided by
non-VRE resources to cover the demand. This means that one D~ ðtÞ ¼ Zin D þ ðtÞZ1
out D ðtÞ, ð6Þ
would hope for EB þ g ¼ 1 for g r 1 and EB ¼0 for g 41, which can
be rewritten as and Hðt min 1Þ ¼ H0 is an initial value to be determined, t min being
the first hour of the time series. The storage works in the
EB ¼ ð1gÞ þ , ð4Þ
following way: any excess power at any time is fed into the
3
because otherwise, more energy is put into the system than is storage with an efficiency of Zin , unless the storage size is
demanded. This also means that, with a perfect storage, or with exceeded, in which case the storage is full. Any deficits in VRE
perfect demand side management, (4) would hold (in fact, this power generation as compared to the load are covered by the
could be the definition of being perfect). storage with an efficiency of Zout , except if the storage runs empty,
Unfortunately, (4) cannot be expected to hold. This means that, in which case the storage only provides partial coverage of the
for a given g, we could have EB 4 ð1gÞ þ , meaning that on deficit. The dispatch strategy is illustrated in Fig. 2.
average, we need additional balancing. This leads us to define To ensure storage neutrality, i.e. that the storage provides
the additional average balancing Eadd
B as only as much energy as is stored, we determine H0 in the
following way: first, it is examined whether the generation
Eadd
B ¼ EB ð1gÞ þ : can match the demand. This is the case if D ~ sums up to a non-
P ~
One can think of Eadd
B as a measure of how well the VRE resources negative value, t D ðtÞ Z 0 (equality is called equilibrium) or
can be integrated into the power system, or how much energy equivalently
one on average would need to move in time to get perfect X X
integration. We stress that Eadd is a property that has to do with Zin Zout D þ ðtÞ Z D ðtÞ: ð7Þ
B t t
an average—it is not a priori possible to assign a time series to the
additional balancing. As one always – given g – can find EB from If this is the case, we set the temporary variable H00 ¼ C S , if not,
Eadd
B by adding ð1gÞ þ , and because Eadd B makes some of the H00 ¼ 0. H00 is then used as an initial guess for H0, and the final
definitions related to storages easier, we will mainly use Eadd B storage filling level time series value Hðt max Þ, t max being the last
rather than EB, except when we deal with lossy storages, as the hour of the time series, is found. We then choose H0 ¼ Hðt max Þ.
concept of additional balancing in this case becomes more With this choice, the storage becomes neutral, which means that
complicated.
2.4. Storage
the average demand. Thus, this scenario has a high over-all going from g ¼ 1:03 to g ¼ 1:12, be it measured in environmental
energy efficiency and a low amount of overinstallation is needed. impact, economical terms or otherwise. If a highly efficient short-
As such it is an example of a fully renewable European electricity term storage comparable to our 6-h storage does not become
supply based on wind, solar PV and a conservative estimate of feasible, the impact of a hydrogen storage is much greater, as it
hydro power generation alone. reduces the needed g from 1.52 to 1.15 or 1.09, depending on
whether the storage is limited to 25 TWh or not.
In conclusion, we find a significant synergy between storage
7. Discussion and conclusion and balancing. In particular, the effect of a highly efficient short-
term storage dramatically reduces the balancing energy needs
We have investigated the ability of the power system to and allows for efficient use of a mix close to the seasonal optimal
integrate any level of wind and solar power in a fully intercon- mix of wind and solar power. However, the additional gain by
nected European power system. In our analysis, we have focused increasing the storage size of such a storage further is limited.
on the ability of both short-term and seasonal storage systems to A seasonal storage will most likely have a low efficiency, but in
increase the integrability. Our study addresses how the introduc- combination with a highly efficient short-term storage, the
tion of a certain amount of storage influences (1) the balancing efficiency of the combined storage system is increased. This
energy needs and (2) the optimal generation mix. The analysis is reduces the needed amount of overproduction in a fully renew-
based on high-resolution hourly weather data and electricity load able scenario. With a balancing of only 150 TWh=year, e.g. coming
for 27 European countries over the 8-year period 2000–2007. from biomass or hydro power, a highly efficient 6-h storage and a
Without additional balancing, a very pronounced cusp singu- 25 TWh hydrogen storage employed, we find that a fully renew-
larity of the needed storage size appears, peaking where the able scenario can be realized with an average wind and solar
system is fully renewable (c.f. Figs. 5 and 8b). This is due to the power production of only 3% more than the average load.
fact that the storage filling level time series changes from being Increasing the hydrogen storage size further does not lower the
almost empty most of the time to being almost full most of the time, needed amount of overproduction substantially, and increasing
resembling a phase-transition. At its peak, the storage size amounts to the figure to 12%, a hydrogen storage can be completely avoided.
at least 320 TWh or 10% of the annual load. For hydrogen storage and At present, the intracontinental power grid is becoming a
with suboptimal mix, the figure is higher. The technical implementa- bottleneck for the integration of non-dispatchable renewables
tion and physical location of such a storage is severely limited and such as wind and solar power. We find that even with a perfect
may not be feasible (Heide et al., 2010). With additional balancing, grid, a combined penetration of wind and solar power of about
the need for such a large storage can, however, be avoided. 50% will lead to the need for an energy storage in order to avoid
For storages of limited size, we have identified that a storage large losses. Investment in an efficient (virtual or physical)
capable of storing 6 h of average consumption – a quarter of an storage able to store the average demand for 6 h is very important
average day – is enough to remove the largest impact of the as it bears large benefits. However, we find that building a storage
intraday mismatch, and thus significantly reduce the balancing large enough to handle all surplus generation at large penetra-
energy needs (c.f. Fig. 8). This implies that the optimal mix is tions is unfeasible. A moderately large storage provides almost
shifted towards the seasonal optimal mix. A high-efficiency 6-h the same benefits, in particular in combination with the highly
storage more than halves the balancing energy needs at g ¼ 1:00 efficient small storage.
for a mix of aW ¼ 0:60. The impact is even greater for both higher
and lower penetration levels, allowing for full integration up to an
average combined wind and solar power generation of about 75% Acknowledgments
(c.f. Fig. 8a). Without the 6-h storage, this full integration can only
be achieved up to about 50% for a fully connected Europe. We would like to thank Rolando A. Rodriguez for proofreading
As indicated in Fig. 9, the impact of a 6-h lossless storage (at several earlier versions of this manuscript and the two anon-
optimal mix) is under certain circumstances almost as good as an ymous referees for comments that markedly improved the quality
unlimited hydrogen storage (at optimal mix). As opposed to the of this final paper. The first author was supported by the Carlsberg
320 TWh storage described above, a 6-h storage, which corre- Foundation during all stages of the work. The second author is
sponds to 2.2 TWh, can be realized by combining many different supported by DONG Energy.
technological solutions including physical storages and time-shift
technologies. Our results indicate that political support for the
development of such technologies will be very beneficial for the Appendix A. Strategies
system when the VRE penetration reaches 50% and will continue
to have a great impact even in the fully renewable regime. In this appendix, we discuss the optimality of the ‘‘storage first’’
A seasonal storage can most likely only be realized with a low strategy with regard to balancing minimization for a given storage
round-trip efficiency. A low-efficiency storage with a round-trip size and give a few examples of other strategies with other
efficiency of 0.36 (0.60 each direction) and a storage size of optimality properties. The fact that the ‘‘storage first’’ strategy is
25 TWh (capable of providing up to 0:60 25 TWh ¼ 15 TWh of indeed optimal with respect to minimizing balancing energy follows
stored energy), is technically realizable e.g. as hydrogen stored in from a simple induction argument, which is left to the interested
solution mined salt caverns in northern Germany. In this scenario, reader as an easy exercise for the mathematically trained.
the high-efficiency 6-h storage serves the purpose of reducing the We stress that this strategy is not the only optimal strategy in this
conversion losses, making the 150 TWh=year of balancing suffi- respect. Recall that the presented storage first strategy acts on an
cient for a g of just 1.03 at an aW ¼ 0:56 mix (again, c.f. Fig. 9). hour-by-hour basis using the storage to cover deficits in the
Without the 6-h storage, the corresponding number would be generation–load mismatch time series if the storage is not empty,
g ¼ 1:15 ðaW ¼ 0:80 mix). Two fully renewable scenarios can be and balancing in case the storage is or runs empty, and in case of a
realized with a lossless 6-h storage: one with g ¼ 1:03 and a positive mismatch, the excess generation is put into the storage if the
25 TWh hydrogen storage and one with g ¼ 1:12 and no addi- storage is not already full. This way of acting makes no use of
tional storage. Whether there should be a hydrogen storage forecasts. Assuming perfect forecast, however, the same action could
depends on the price of the 25 TWh compared to the price of be performed on larger time intervals, in our case same-sign
650 M.G. Rasmussen et al. / Energy Policy 51 (2012) 642–651
intervals, i.e. we split the time series in the largest chunks of intervals
where the mismatch time series does not change sign. Having done
this, one can now check to see how much of the integral of a positive
mismatch interval can fit into the storage and how much of the
integral of a negative mismatch interval can be covered by what is in
the storage. Depending on what one wants to reduce (excess,
balancing, storage charge/discharge power quantiles), one can then
either cover the negative mismatches using the storage to cover the
most negative part of the mismatch (trough filling) or using the
storage to cover with a more or less constant storage power output
(top filling), and vice versa for the positive mismatches (peak shaving
and bottom shaving, respectively). In either case, the energy sum of
what is used or absorbed by the storage in any given same-sign
interval remains the same as for the storage first strategy, leaving the
minimization with respect to balancing energy unaffected. The
original version (Strategy 1—‘‘storage first’’) and two combinations
(peak shaving–trough filling (Strategy 2) and bottom shaving–top
filling (Strategy 3), respectively) are illustrated in Fig. A1.
Fig. A2. Storage filling level (top panels) and mismatch, storage flow, excess and
balancing power (bottom panels) for Strategies 4 and 5, respectively. See Fig. A1
for explanation of the color coding.
Beaudin, M., Zareipour, H., Schellenberglabe, A., Rosehart, W., 2010. Energy storage Kruse, B., Grinna, S., Buch, C., 2002. Hydrogen—Status og Muligheter. Technical
for mitigating the variability of renewable electricity sources: an updated Report 6. Bellona (in English).
review. Energy for Sustainable Development 14, 302–314. Lehner, B., Czisch, G., Vassolo, S., 2005. The impact of global change on the
Czisch, G., 2005. Szenarien zur zukü nftigen Stromversorgung, kostenoptimierte hydropower potential of Europe: a model based analysis. Energy Policy 33,
Variationen zur Versorgung Europas und seiner Nachbarn mit Strom aus 839–855.
erneuerbaren Energien. Ph.D. Thesis. Universität Kassel. Lund, H., Mathiesen, B.V., 2009. Energy system analysis of 100% renewable energy
Czisch, G., Giebel, G., 2007. Realisable scenarios for a future electricity supply systems—the case of Denmark in years 2030 and 2050. Energy 34, 524–531.
based 100% on renewable energies. In: Energy Solutions for Sustainable Martinot, E., Dienst, C., Weiliang, L., Qimin, C., 2007. Renewable energy futures:
Development Proceedings Risø International Energy Conference, Risø National targets, scenarios, and pathways. Annual Review of Environment and
Library, Technical University of Denmark, pp. 186–195. Resources 32, 205–239.
ECF, 2010. Roadmap 2050: A Practical Guide to a Prosperous, Low-carbon Europe. Rodriguez, R., Andresen, G.B., Becker, S., Greiner, M., 2012. Transmission needs in a
Report. European Climate Foundation.
fully renewable pan-European electricity system. In: Uyar, T.S., Sağlam, M.,
Gilhaus, A., 2007. Natural Gas Storage in Salt Caverns—Present Status, Develop-
Sulukan, E. (Eds.), 2nd International 100% Renewable Energy Conference and
ments and Future Trends in Europe. Technical Conference Paper. Solution
Exhibition (IRENEC 2012) Proceedings, Maltepe, Istanbul, Turkey, pp. 320–324.
Mining Research Institute.
Schaber, K., Steinke, F., Hamacher, T., 2012a. Transmission grid extensions for the
Hedegaard, K., Meibom, P., 2012. Wind power impacts and electricity storage—a
integration of variable renewable energies in Europe: who benefits where?.
time scale perspective. Renewable Energy 37, 318–324.
Heide, D., von Bremen, L., Greiner, M., Hoffmann, C., Speckmann, M., Bofinger, S., Energy Policy 43, 123–135.
2010. Seasonal optimal mix of wind and solar power in a future, highly Schaber, K., Steinke, F., Mühlich, P., Hamacher, T., 2012b. Parametric study of
renewable Europe. Renewable Energy 35, 2483–2489. variable renewable energy integration in Europe: advantages and costs of
Heide, D., Greiner, M., von Bremen, L., Hoffmann, C., 2011. Reduced storage and transmission grid extensions. Energy Policy 42, 498–508.
balancing needs in a fully renewable European power system with excess Stone, H.B.J., Veldhuis, I., Richardson, R.N., 2009. Underground hydrogen storage in
wind and solar power generation. Renewable Energy 36, 2515–2523. the UK. The Geological Society, London, Special Publications 313, 217–226.
Jacobson, M.Z., Delucchi, M.A., 2011. Providing all global energy with wind, water, Strbac, G., 2008. Demand side management: benefits and challenges. Energy Policy
and solar power, Part I: technologies, energy resources, quantities and areas of 36.
infrastructure, and materials. Energy Policy 39, 1154–1169. Widén, J., 2011. Correlations between large-scale solar and wind power in a future
Kempton, W., Pimenta, F.M., Veron, D.E., Colle, B.A., 2010. Electric power from scenario for Sweden. IEEE Transactions on Sustainable Energy 2, 177–184.
offshore wind via synoptic scale interconnection. Proceedings of the National Yao, R., Steemers, K., 2005. A method of formulating energy load profile for
Academy of the Sciences of the United States of America 107, 7240–7245. domestic buildings in the UK. Energy and Buildings 37, 663–671.